BIODIVERSITYRESEARCH

Conservation implications of omittingnarrow-ranging taxa from speciesdistribution models, now and in thefuturePhilip J. Platts1†*, Raquel A. Garcia2,3,4†, Christian Hof5, Wendy Foden6,7,

Louis A. Hansen2, Carsten Rahbek2,8 and Neil D. Burgess2,9,10†

1Environment Department, University of

York, Heslington, YO10 5DD York, UK,2Center for Macroecology, Evolution and

Climate, Natural History Museum of

Denmark, University of Copenhagen,

Copenhagen, Denmark, 3Department of

Biodiversity and Evolutionary Biology,

National Museum of Natural Sciences,

CSIC, Calle Jos�e Gutierrez Abascal, 28006

Madrid, Spain, 4Rui Nabeiro Biodiversity

Chair, University of �Evora, CIBIO, Largo

dos Colegiais, 7000 �Evora, Portugal,5Biodiversity and Climate Research Centre

(BiK-F) & Senckenberg Gesellschaft f€ur

Naturforschung, Senckenberganlage 25,

60325 Frankfurt, Germany, 6School of

Animal, Plant and Environmental Sciences,

University of the Witwatersrand, 2050

Johannesburg, South Africa, 7Climate

Change Specialist Group, IUCN Species

Survival Commission, Gland, Switzerland,8Department of Life Sciences, Faculty of

Natural Sciences, Imperial College London,

Silwood Park, Buckhurst Road, Ascot, SL5

7PY Berkshire, UK, 9United Nations

Environment Programme World

Conservation Monitoring Centre, 219

Huntingdon Road, CB3 0DL Cambridge,

UK, 10World Wildlife Fund–US, 1250 24th

Street NW, Washington, 20037 DC, USA

†Contributed equally.

*Correspondence: Philip J. Platts,

Environment Department, University of York,

Heslington, York YO10 5DD, UK.

E-mail: philip.platts@york.ac.uk

ABSTRACT

Aim Species distribution modelling (SDM) is commonly used to predict spatial

patterns of biodiversity across sets of taxa with sufficient distributional records,

while omitting narrow-ranging species due to statistical constraints. We investi-

gate the implications of this dichotomy for conservation priority setting in

Africa, now and in the future.

Location Sub-Saharan Africa (excluding islands).

Methods We use multivariate ordination to characterize climatic niches of 733

African amphibians, distinguishing between species eligible for large-scale cor-

relative SDM (≥ 10 records at 1° resolution) and those omitted due to insuffi-

cient records. Species distributions are projected under current and future

climates using simple niche envelopes. Empirical priorities are derived sepa-

rately on the eligible and omitted sets and compared with three existing large-

scale conservation schemes.

Results Of the 733 amphibian species, 400 have too few records for correlative

SDM, including 92% of those threatened with extinction (VU/EN/CR). Omit-

ted species typically occupy topographically complex areas with cooler, wetter

and less seasonal climates, which are projected to experience lower rates of cli-

matic change. Priorities derived from omitted species have greater congruence

with existing conservation schemes. Under future climate, priorities for eligible

species shift towards those for omitted species. Similarly, while omitted species

often lose climate space at 1° resolution, persistent populations tend to coin-

cide with existing conservation schemes.

Main conclusions Under current climate, statistical restrictions on SDM sys-

tematically downplay important sites for narrow-ranging and threatened spe-

cies. This issue spans taxonomic groups and is only partially mitigated by

modelling at finer scales. Effective biodiversity conservation, now and in the

future, relies on our capacity to project geographic determinants of all species,

and thus, a wider range of approaches is essential. We conclude, however, that

future persistence among narrow- and wide-ranging species alike will be highest

within sites already identified for conservation investment and that the focus

on these sites ought to be maintained.

Keywords

Africa, amphibians, biodiversity, climate change, conservation priorities, threa-

tened species.

DOI: 10.1111/ddi.12244ª 2014 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/ddi 1307

Diversity and Distributions, (Diversity Distrib.) (2014) 20, 1307–1320A

Jou

rnal

of

Cons

erva

tion

Bio

geog

raph

yD

iver

sity

and

Dis

trib

utio

ns

INTRODUCTION

Limited funds for tackling unprecedented rates of biodiver-

sity loss dictate that spatial priorities for conservation must

be identified and periodically revised (e.g. Margules & Pres-

sey, 2000; Butchart et al., 2010). In general, biologically rich

areas with high irreplaceability, which are also vulnerable to

degradation and loss, are given the highest ranks (e.g. Statt-

ersfield et al., 1998; Myers et al., 2000; Wilson et al., 2006).

In recent years, the predicted (modelled) impact of climate

change on species distributions has become an additional

consideration for conservation priority setting (Hannah

et al., 2002; Huntley et al., 2006; Dawson et al., 2011).

The collection of biological data is time-consuming and

expensive, particularly for rare species (Ahrends et al., 2011).

Despite its importance for conservation, such work is chroni-

cally underfunded (Balmford & Gaston, 1999; Da Fonseca

et al., 2000). In prioritizing sites for conservation, the near-

term solution is to project existing distributional data across

space and, increasingly, into the future, using environmental

correlates (Pressey et al., 2000). Correlative models that pre-

dict species distributions under climate change are frequently

applied in the scientific literature and are widely cited by

conservation planners, who seek to ensure that priorities

remain valid under future climates (Williams et al., 2005;

Hole et al., 2009; Shoo et al., 2011).

Considerable literature exists on species distribution mod-

elling (SDM), its strengths and limitations (e.g. Wiens et al.,

2009; Ara�ujo & Peterson, 2012; Franklin, 2013). For conserva-

tion prioritization, perhaps the most fundamental constraint

is that, for many SDM methods, only species with adequate

numbers of spatially distinct occurrence records can be mod-

elled (Stockwell & Peterson, 2002). In tropical regions, which

contain most of the world’s biodiversity, species are more

often narrow ranging than in temperate zones (Wiens et al.,

2006) and even common species’ distributions tend to be less

well documented (Da Fonseca et al., 2000; K€uper et al., 2006;

Feeley & Silman, 2011). Consequently, species that have tradi-

tionally been used to define conservation priorities are the

ones most likely to be omitted from correlative SDM. An

important question for conservation planning is whether this

omission of narrow-ranging species results in a systematic

bias, now or in the future, towards or away from sites priori-

tized under existing conservation schemes.

We address this question using African amphibians as an

exemplar group. Amphibians contain many narrow-ranging

and threatened species, but also wide-ranging species more

typical of higher latitudes. A further motivation for studying

amphibians is the high rates of threat they are predicted to

face from climate change, habitat loss and disease, especially

in Africa (Sodhi et al., 2008; Hof et al., 2011a; Foden et al.,

2013). Comparing species eligible for correlative SDM with

those most often omitted due to statistical constraints, we

aim to establish whether this dichotomy: (1) restricts SDM

to an unrepresentative sample of present-day conditions and/

or exposure to climate change and (2) has potential to

undermine or reinforce existing conservation priority

schemes, now or in the future. We consider three schemes

that have been widely used by international organizations to

allocate conservation funds: Conservation International’s

Hotspots, BirdLife International’s Endemic Bird Areas and

the World Wildlife Fund’s Global 200 set of ecoregions. The

extent to which our results generalize across spatial resolu-

tions and taxonomic groups is discussed, and avenues for

uniting established conservation prioritization procedures

with SDM projections under climate change are identified.

METHODS

Amphibian data

Distributional data were collated for 790 species of amphib-

ian found only on mainland Africa south of the Sahara (Hof

et al., 2011a, with database updated to February 2014). The

taxonomy of this database was matched to the IUCN list of

amphibians for the same region (www.iucnredlist.org,

accessed February 2014). The IUCN list includes 747

amphibian species, of which 733 were matched to our data-

base criteria (Table S1 in Supporting Information). The spe-

cies data are reliable to 1° resolution (111 km at the

equator), consistent with most studies of continental-scale

biodiversity pattern and prioritization under climate change

(e.g. McClean et al., 2005; Huntley et al., 2006; Hole et al.,

2009; Garcia et al., 2012).

Each amphibian was allocated to one of two species sets:

those that are eligible for correlative SDM at 1° resolution

(N = 333) and those that are not (N = 400), hereafter

referred to, respectively, as ‘eligible’ and ‘omitted’ amphibi-

ans (Fig. 1 and Table S1). We used a ten-record cut-off for

the minimum number of records required for SDM. Stricter

restrictions have been suggested, such as 50 records in total

or ten records per predictor variable in the model (Harrell

et al., 1984; Stockwell & Peterson, 2002; Wisz et al., 2008).

Depending on the threshold (10–50 records), 33–46% of

omitted amphibians would be regarded as ‘rare’ according to

standard definitions in conservation biology (i.e. lower quar-

tile of range sizes within the group; Gaston, 1994). For both

species sets, we identified the proportion of species listed as

Threatened on The IUCN Red List under different threat

categories (Tables S1 and S2).

Climate data

We selected four, weakly collinear (Pearson’s |r| < 0.7; Dor-

mann et al., 2013), climatic predictor variables, which pro-

vide an equal balance between temperature and rainfall and

between annual and seasonal conditions: mean annual tem-

perature, annual temperature range, annual rainfall and rain-

fall seasonality. The latter reflects rainfall contrasts across

seasons (Rs) in relation to the total annual rainfall (RT):

∑s = 1..4 |Rs – RT/4|/RT (Walsh & Lawler, 1981).

1308 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.

Climate grids for the baseline period (1950–2000) were

from WorldClim (Hijmans et al., 2005), using mean values

to resample from 30″ (1 km) to the 1° resolution of the

species data. For future climate (2071–2100), we used a

multimodel average over eight CMIP5 GCMs (CanESM2,

CNRM-CM5, EC-EARTH, GFDL-ESM2G, HadGEM2-ES,

MIROC5, MPI-ESM-LR and NorESM1-M). Each GCM was

dynamically downscaled to 0.44° (c. 50 km) using the

SMHI-RCA4 regional climate model, debiased against

WorldClim baselines (Platts et al., 2014) and then resam-

pled to the 1° resolution of the species data. Two represen-

tative concentration pathways of the IPCC-AR5 were

considered: RCP4.5 and RCP8.5, which here project mean

temperature anomalies across sub-Saharan Africa of, respec-

tively, 2.6 �C and 4.5 �C by 2071-2100 relative to the

WorldClim baseline.

Niche differences between species

Using multivariate ordination (Outlying Mean Index, OMI;

Dol�edec et al., 2000), we assessed differences between

observed distributions for eligible and omitted species in

terms of (1) climatic niche space and (2) projected exposure

to change (future climate anomalies). The OMI identifies

ordination axes that optimize separation between species

occurrences and quantifies niche position and niche breadth

for each species along those axes. Niche position, or margin-

ality, is measured as the deviation of the mean climatic con-

ditions – or future climate anomalies – occupied by a

species (its centroid) from the mean conditions/anomalies of

the study region (origin of the axes). The larger the devia-

tion, the more marginal the species’ niche position is in rela-

tion to the study region as a whole. Niche breadth, or

tolerance, is quantified as the dispersion of climatic condi-

tions/anomalies occupied by species, with larger dispersion

indicating higher tolerance or more variable exposure to

change.

Anomalies for temperature variables were computed by

subtracting the future from the present values; anomalies for

rainfall variables were given by the ratio of future to present

values (Fig. S2 in Supporting Information). For each OMI

analysis, we performed randomization tests (1000 permuta-

tions) to determine whether species’ niche positions or expo-

sure to change) differed significantly from random

expectations. Calculations were performed in R using the

ade4 package (Chessel et al., 2004; R-Core-Team, 2012).

(a)

(b)

Figure 1 (a) Mountainous areas in mainland Africa south of the Sahara (Kapos et al., 2000), overlaid with the distribution of records

for 400 narrow-ranging amphibians (< 10 records at 1° resolution) and 333 more widespread (≥ 10 records) amphibians, the latter

being eligible for species distribution modelling, while the former are omitted due to statistical constraints. Stacked bar chart (b) details

the number of distinct 1° grid cells known to be occupied by African amphibians, including a breakdown of threat in each range-size

category, according to The IUCN Red List (www.iucnredlist.org, accessed February 2014).

Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd 1309

Species distribution models and conservation priorities

Constructing SDMs without occurrence thresholds

Amphibian distributions were projected in space and time

using multidimensional niche envelopes (MDNE). Popular

SDM methods, such as maximum entropy and statistical

regression, require sufficient occurrence records to paramete-

rise a species’ response to, and relative importance of, pre-

dictors in a model. The MDNE more simply classifies all

conditions within a species’ observed climatic range as uni-

formly viable, and conditions beyond as wholly unsuitable,

with all predictors given equal weight (Busby, 1991; Walker

& Cocks, 1991). This approach requires no occurrence

threshold and so can be applied consistently across the eligi-

ble and omitted species sets.

Rather than using mean predictor values at 1° resolution

to define the MDNE, we used the lower and upper quartiles

of climate conditions within 1° cells, resampled from 30″.

This definition reduced the number of type II errors (false

negatives) that would otherwise arise from topographic com-

plexity within 1° cells. It also has the advantage of being

defined for species having only one record of occurrence,

their MDNE being the interquartile range of conditions

within the documented 1° cell.MDNE projections could not be validated directly: by defi-

nition, there can be no errors of omission against the 1°training data, while specificity is unknown. We could, how-

ever, gauge comparability of MDNE projections for eligible

species with those obtained from MAXENT (ver. 3.3.3e; Phillips

& Dudik, 2008), using the latter as test data to validate the

MDNE results. MAXENT was chosen for its widespread appli-

cation and favourable performance in comparative studies

(e.g. Hernandez et al., 2006). Ten replicates were performed

for each eligible species, withholding distinct tenths of the

presence data for testing. The final prediction for a given

pixel was the mean value over this tenfold cross-validation.

Continuous occurrence probabilities were dichotomized into

presence–absence maps by maximizing the sum of sensitivity

and specificity (Liu et al., 2013).

Future predictions were constrained under a no-dispersal

scenario, which is justified by the coarse resolution of the

grid (Bateman et al., 2013). Climate change velocities at 1°resolution will likely outpace dispersal capabilities for most

African amphibians: to traverse a single pixel, the leading

edge of a population would need to migrate 1 km/y, subject

to continuous corridors for dispersal and a steady change in

climate. Yet, temporal climatic variability and, for all but the

most generalist of species, habitat fragmentation are likely to

curtail future rates and patterns of range shift relative to the

past (Early & Sax, 2011; Hof et al., 2011b; Bennie et al.,

2013; Pyron & Wiens, 2013).

Conservation priorities

Grouping predicted distributions by species set (eligible or

omitted) and time period (present or future), we derived

empirical priority metrics for spatial comparison with three

large-scale conservation priority schemes: Biodiversity Hot-

spots (Conservation International; Mittermeier et al., 2004),

Endemic Bird Areas (BirdLife International; Stattersfield

et al., 1998) and the Global 200 set of ecoregions (World

Wildlife Fund US; Olson & Dinerstein, 1998).

Empirical metrics were chosen to complement these

schemes, without confounding the intended comparison

between projections based on narrow- versus wide-ranging

species (e.g. by explicit inclusion of data on vulnerability

and habitat; Margules & Pressey, 2000; Williams et al.,

2005; Moilanen et al., 2009). The metrics were: (1) the 100

grid cells with the highest species richness; (2) the 100 grid

cells with the highest range-size rarity, defined as the sum

of inverse range sizes of all species in a grid cell; and (3)

minimum sets of grid cells that represent all species at

least once (greedy complementarity). To identify minimum

sets, we began with a full set of non-empty cells and

reduced it, one cell at a time, until no further reduction

was possible without excluding one or more species. The

procedure was randomized and repeated 1000 times. The

final priority metric was the union over all optimal solu-

tions. Congruence with existing schemes was defined by

the number (richness and rarity) or percentage (minimum

sets) of empirically prioritized cells that fell within existing

schemes.

RESULTS

Degree of (non-climatic) threat

Of the 733 amphibian species, 191 are designated Threatened

on The IUCN Red List: 57 (8%) are vulnerable, 96 (13%)

are endangered, 37 (5%) are critically endangered, and one

was extinct in the wild (Nectophrynoidesasperginis, now rein-

troduced to original locality). Of the remainder, 40 (5%) are

designated near threatened and 354 (48%) least concern,

while 148 (20%) are data deficient. Because restricted range

is one of the criteria for Red Listing species, the degree of

threat is highly skewed towards taxa with fewer occurrence

records at the 1° model resolution (Fig. 1b). Only 16 of the

191 threatened amphibian species are eligible for correlative

SDM, whereas the remaining 175 threatened species, includ-

ing nearly all of those in the highest threat categories

(endangered and critically endangered), are too narrow rang-

ing to model (Tables S1 and S2).

Niche differences between species

Omitted amphibians have been recorded predominantly in

tropical highlands (regions 1–9 in Fig. 1a), in grid cells with

significantly higher proportions of mountain area (as per

Fig. 1a, Wilcoxon signed-rank one-sided, P < 0.05) com-

pared with eligible species. Further hotspots for omitted spe-

cies include the Cape Fynbos and Drakensberg of South

Africa (regions 10–11). By contrast, at least one eligible

amphibian has been recorded in the majority of grid cells

across sub-Saharan mainland Africa (Fig. 1a).

1310 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.

Current conditions

The first two ordination axes of the OMI analysis explained

94% of niche separation between species, with the first axis

capturing rainfall gradients and temperature range and the

second loaded mainly with mean temperature (Table S3).

Eligible and omitted species were found to be significantly

different in both their marginality and tolerance (Fig. 2c;

Wilcoxon signed-rank one-sided, P < 0.05). Whereas eligible

species are more ubiquitous on the study area, omitted spe-

cies occupy climatic conditions that deviate from the average

conditions for mainland Africa south of the Sahara (Fig. 2a–

b). These marginal environments are characterized by higher

annual rainfall with lower seasonality (axis 1) and by cooler

temperatures (axis 2). Omitted species also have climatic

niches that are narrower than those of eligible species. These

patterns hold when the analysis is restricted to species with

statistically significant values of marginality (Fig. S1a).

Exposure to change

The OMI analysis on climate anomalies also revealed contrast-

ing patterns between eligible and omitted species (Fig. 2e-i).

Under both emission pathways, niche differentiation was

explained by anomalies in mean annual temperature and, to a

lesser extent, annual temperature range and annual rainfall

(axis 1). Axis 2 captured precipitation gradients but with lower

loadings. Together, the two axes explained 76–78% of the total

variance in species’ exposure to climate change, depending on

(a) (b) (c)

(d) (e) (f)

(i)(h)(g)

Figure 2 Niche differences between species that are eligible for SDM and the omitted species set. Scatterplots show species centroids in

environmental space, represented by the first two ordination axes of the Outlying Mean Index, for present climate (a–b) and for climate

anomalies under emission pathways RCP4.5 (d–e) and RCP8.5 (g-h). Boxplots (c, f, i) compare marginality and tolerance for both sets

of species. Whiskers extend up to 1.5 times the interquartile range from each box. As indicated by non-overlapping notches, omitted

species are significantly more marginal and less tolerant than eligible species (Wilcoxon signed-rank, P < 0.05; Fig. S1).

Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd 1311

Species distribution models and conservation priorities

the emission pathway (Table S3). Omitted species were found

to be significantly more marginal and less tolerant than eligible

species (Fig. 2f,i; Wilcoxon signed-rank one-sided, P < 0.05),

occupying areas projected to be less exposed to warming

(Fig. 2d–e and 2g–h). The same patterns hold when the com-

parison on both OMI axes is restricted to species with signifi-

cant marginality (Fig. S1b–c).

Conservation priorities

Model performance

The predictive performance of MAXENT models for eligible

species was good, with median cross-validated AUC of 0.95

(interquartile range, IQR: [0.91, 0.97]). At the presence–

absence threshold, the median sensitivity was 0.94 [0.91,

0.96] and specificity was 0.91 [0.88, 0.94]. The degree to

which MAXENT extrapolated beyond the parameter range of

training data was minimal, except for novel high tempera-

tures in the Sahelian zone (Fig. S2). The most important pre-

dictor was annual rainfall (median contribution: 36%, IQR:

[20, 50]), followed by mean annual temperature (26%, [6,

42]), annual temperature range (23%, [6, 35]) and rainfall

seasonality (15%, [5, 19]; Fig. S3).

Using MAXENT predictions as test data, MDNE maps for

eligible species had median sensitivity of 0.92 (IQR: [0.78,

0.99]) and median specificity of 0.91 [0.79, 0.97]. For species

richness and range-size rarity, MDNE and MAXENT results

were highly correlated (Spearman’s q = 0.90 and q = 0.87;

Fig. S4).

Priority metrics for eligible versus omitted species

Under present climate, around a third of priority cells identi-

fied by the species richness and range-size rarity metrics were

common to both the eligible and omitted species sets

(Table 1). Discrepancies were highest in the Ethiopian

Highlands, Great Rift Valley, Eastern Arc Mountains and

Coastal Forests of East Africa, and in the Cape Fynbos and

Drakensberg of South Africa (Fig. 3a–b), which were all lar-

gely absent from eligible species’ priorities. In the future, pri-

orities for eligible and omitted species converged under both

metrics, increasing to 70–81 cells (of a possible 100) for spe-

cies richness, depending on the emission pathway, and to

55–72 cells for range-size rarity (Table 1).

The current agreement between the minimum set metric

for eligible (62 cells) and omitted (195 cells) species was low

(9 cells in common; Table 1). In particular, the Fouta Djal-

lon/Western Upper Guinea Forests, Angola Highlands, Ethio-

pian Highlands, Eastern Zimbabwe Mountains, Eastern Arc

and Coastal Forests were under-represented by the eligible

species set (Fig. 3c). In the future, the size of minimum sets

reduced for omitted species, due to many narrow-ranging

amphibians losing all suitable climate space at the 1° resolu-

tion (Fig. 4). Agreement between eligible and omitted species

reduced slightly under RCP4.5, but doubled to 18 cells under

RCP8.5 (Table 1).

Comparison with existing conservation schemes

Congruence between empirical priority metrics and existing

schemes was generally higher on the omitted species set than

on the eligible species set, although this varied depending on

the region and metric considered (Fig. 4). For omitted

amphibians, 39–70 (richness) and 32–74 (range-size rarity)

of prioritized cells coincided with existing schemes, com-

pared with eligible species congruence of 13–50 and 14–46

cells, respectively. Congruence according to minimum sets

was more similar for omitted versus eligible species (44–64%

vs. 32–71%), suggesting that this metric may be more robust

to the omission of narrow-ranging species from priority

assessments (Figs. 3 & 4). For both species sets, the degree of

congruence with the Global 200 was often higher than with

Biodiversity Hotspots, which was higher than with Endemic

Bird Areas, due to differences in overall coverage between

these schemes (Fig. 4d).

Projecting empirical metrics under future climate, congru-

ence with existing schemes reduced in western Africa while

generally increasing in eastern and southern Africa. Overall,

priorities for eligible species were projected to shift towards

existing schemes, due to greater climatic stability at these

Table 1 Number of grid cells common to spatial priority

metrics for eligible and omitted amphibians, now and in the

future. Eligible species can be modelled using popular

methodologies for species distribution modelling (≥ 10 records),

while omitted species cannot. Congruence increases in the

future, especially under IPCC-AR5 emission pathway RCP8.5

Priority metric

Present

(1950–2000)

RCP4.5

(2071–2100)

RCP8.5

(2071–2100)

Species richness 35 70 81

Range-size rarity 33 55 72

Minimum sets 9 6 18

Figure 3 Conservation priorities derived empirically from data on amphibian distributions, separately for species typically omitted

from SDM (< 10 records) and those that are eligible (≥ 10 records). Spatial distributions were projected using multidimensional niche

envelopes. Future climate was simulated by a regional climate model (eight driving GCMs) under two representative concentration

pathways of IPCC-AR5. For species richness (a) and range-size rarity (b), the top 100 scoring cells are selected as priority sites.

Minimum sets (c) identify the smallest possible subsets of cells that represent all species at least once; where the solution is not unique,

the union across solutions is mapped. Ellipses highlight key sites under-represented on the eligible species set. For a key to regions, see

Fig. 1a.

1312 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.

(a)

(b)

(c)

Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd 1313

Species distribution models and conservation priorities

sites (Fig. S2). Similarly, while omitted species frequently lost

all climate space at the 1° resolution, persistent populations

tended to coincide with existing schemes (Fig. 4).

DISCUSSION

Natural habitats are changing at a rate unprecedented in

human history. Faced with finite resources for addressing the

decline of global biodiversity, it is necessary to locate priority

sites for conservation investment (e.g. Olson & Dinerstein,

1998; Stattersfield et al., 1998; Mittermeier et al., 2004).

Without global priorities, resources would be spread too

thinly to have any meaningful impact on species conserva-

tion (Pimm et al., 2001; Wilson et al., 2006). Large-scale pri-

ority schemes, such as Biodiversity Hotspots, Endemic Bird

Areas and Global 200 ecoregions, were developed directly

from the available data on species occurrence (particularly

local endemism) and threats, augmented by expert opinion.

More recently, species distribution modelling (SDM) has

become an almost ubiquitous technique in the conservation

scientist’s toolbox, accelerated in popularity by free Internet

access to large databases of herbaria and zoological speci-

mens (Graham et al., 2004), digitized environmental data

and a latent demand for predictive spatial models, particu-

larly with respect to the impacts of anthropogenic climate

change.

From simple envelope approaches, which employ concise

rule sets to define species range boundaries (e.g. MDNE as

employed here; BIOCLIM, Busby, 1991; HABITAT, Walker

& Cocks, 1991; GARP, Stockwell & Peters, 1999), more

complex modelling procedures have emerged (e.g. MARS

and BRT, Friedman, 1991, 2001; MAXENT, Phillips & Dudik,

2008). Due to their superior capacity to fit the observed

species data, housed within user-friendly desktop applica-

tions, such advances have been widely advocated (Elith

et al., 2006; Franklin, 2010). However, as the authors of

these tools are the first to point out, model calibration

must be underpinned by sufficient data on a species’ distri-

bution to avoid spurious predictions – a prerequisite that

is, almost by definition, not fulfilled by many taxa of high-

est conservation concern, particularly at the coarse spatial

resolutions dictated by species and climate data at global or

continental scales.

In this study, we have highlighted that over half of the

amphibian species in sub-Saharan Africa (excluding islands)

have ranges too narrow for large-scale correlative SDM and

that these species are disproportionately vulnerable, endan-

gered or critically endangered on The IUCN Red List, with

92% of all such species being ineligible for modelling. This is

to be expected, given that range-size is a criterion in the Red

Listing process, but the consequences have rarely been

emphasized. Crucially for studies that infer conservation pri-

orities from SDM, we find that the omitted species occupy

significantly different niche space to eligible species: their

observed distributions are characterized by higher annual

rainfall with lower rainfall seasonality, by cooler and less sea-

sonal temperatures and by more complex topography. This

is consistent with previous studies (De Klerk et al., 2002; Jetz

& Rahbek, 2002; Ohlem€uller et al., 2008; Sandel et al., 2011;

Pyron & Wiens, 2013), which find that climatically and

topographically diverse/distinct areas contain a dispropor-

tionate richness of narrow-ranging species.

Our results show that continental-scale correlative SDM is

restricted to an unrepresentative sample of both current cli-

mate and future climate anomalies, and that this bias is coun-

ter to more traditional frameworks for identifying priority

sites for conservation, which focus precisely on those taxa

most likely to be omitted. Reverting to a simpler form of

SDM – multidimensional niche envelopes – we find that con-

servation priorities derived empirically from the omitted spe-

cies set generally exhibit higher congruence with existing

priority schemes than those inferred from eligible species. In

the future, sites occupied by omitted species are projected

lower rates of warming. Many such sites, especially in eastern

Africa, are projected to become increasingly important for the

persistence of omitted and eligible species alike. Similarly at

the ecoregion level, Iwamura et al. (2013) found that overlap

between current and future climate space is highest within

known centres of endemism, and Sandel et al. (2011) have

shown that low climate velocities during the Late Quaternary

coincide with contemporary hotspots for narrow-ranging

birds, mammals and amphibians. These patterns are consis-

tent with the theory that long-term climatic stability, com-

bined with topographic complexity and geographic isolation,

gives rise to, and may continue supporting, high rates of rar-

ity and endemism (Fjelds�a et al., 1997; Platts et al., 2013).

The projected future convergence of priority sites for nar-

row- and wide-ranging amphibians may therefore echo his-

torical processes underlying contemporary biodiversity

patterns. The 1° resolution used here is, however, too coarse

to capture the variations in local climate, biotic interaction

and adaptive potential that are ultimately central to species

persistence under climate change (Hof et al., 2010). Trivedi

et al. (2008) found that coarse resolution models of high-

altitude plants in Europe overestimated their thermal toler-

ance and thus underestimated mountain-top extinction.

Conversely, Randin et al. (2009) found that topoclimatic

refugia were better captured at finer scales, facilitating species

Figure 4 Spatial congruence between empirically derived amphibian priorities and three existing conservation priority schemes in

western (a), eastern (b) and southern (c) Africa south of the Sahara, according to the partitions and schemes mapped in (d). Bar

heights are the number (richness and rarity metrics) or percentage (minimum sets) of cells that coincide with existing schemes. Bar

widths are proportional to the number of species still with suitable climate space by 2071–2100, under two emission pathways of IPCC-

AR5. Underlying species distributions were mapped on 1° grids using multidimensional niche envelopes.

1314 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.

(a)

(b)

(c)

(d)

Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd 1315

Species distribution models and conservation priorities

persistence (see also Franklin et al., 2013). Our assumption

of no intercell dispersal would need to be revised at finer

scales, because narrow isotherms along altitudinal gradients

facilitate species ability to track changing climates more

locally (Loarie et al., 2009), with the likely effect of further

emphasizing the importance of existing, topographically

complex, priority sites for conservation.

Most fundamentally for our analysis, a finer grid resolu-

tion would disaggregate spatially clustered occurrence

records, increasing the proportion of species eligible for cor-

relative SDM, albeit with increased spatial autocorrelation.

We suggest that the convergence of conservation priorities

for eligible versus omitted species under climate change is

unlikely to be scale dependent in this regard, because sites of

relative climatic stability would remain so, regardless of the

species dichotomy. The effect of spatial scale on present-day,

divergent priority metrics for omitted versus eligible species

is more uncertain and would depend on how the disaggrega-

tion of distributional records varied between the two groups.

A further consideration is that parameterizations for particu-

lar species models are also scale dependent, because the

observed niche is shaped increasingly by non-climatic factors

(e.g. biotic interactions, patch dynamics, habitat quality) at

finer scales (Pearson & Dawson, 2003).

Our ability to investigate these uncertainties is limited by

the fact that much of Earth’s biodiversity is poorly docu-

mented, both in the quantity and in the quality of distribu-

tional records. In Africa, as elsewhere in the tropics, 1° (or

at best 0.5°) grids are the operational default for continental-

scale SDM, because much of the available species data are

not reliably georeferenced at finer resolutions (Graham et al.,

2004; Wiens & Bachelet, 2010; Feeley & Silman, 2011). In

specific regions where high-resolution species data are avail-

able, the issue of threatened species being omitted from

modelling procedures remains pervasive. The Eastern Arc

Mountains, for example, are among the best-documented

sites in Africa (Feeley & Silman, 2011), yet two-thirds of

threatened amphibians have fewer than ten records on a 100

grid (Burgess et al., 2007). Moreover, of the c. 500 plant spe-

cies endemic to the Eastern Arc, none have sufficient herbar-

ium data for correlative SDM on a 1° grid and, even at 30″

resolution, 90% are too narrow ranging to model (Platts

et al., 2013), the majority of these species being assessed as

either threatened or scheduled for Red List assessment

(Gereau et al., 2010).

Novel approaches to address these challenges of spatial

scale and narrow ranges, both within and beyond the field of

SDM, are emerging. Current knowledge on species distribu-

tions can be systematically improved by iterative application

of simple models and targeted fieldwork (Williams et al.,

2009; Platts et al., 2010). Making the most of small sample

sizes, Lomba et al. (2010) combine large numbers of bivari-

ate models, built individually using all pairwise combinations

of predictors. In contrast to a species-level focus, hierarchical

approaches, combining species-specific and community mod-

els (Loarie et al., 2008; Ovaskainen & Soininen, 2011), or

spatial modelling of climatically associated (Golicher et al.,

2008) or community-level (Ferrier & Guisan, 2006) species

pools can help to detect shared patterns of response across

narrow-ranging taxa. Simple measures of site exposure to

different aspects of climate change, which do not rely on

SDM, also provide indications of potential threats and

opportunities for species present in such sites (Williams

et al., 2007; Ohlem€uller et al., 2008; Garcia et al., 2014b).

These methods can be combined with biological trait-

based assessments of species sensitivity (ability to cope in

situ) and adaptive capacity (ability to escape by dispersal or

evolution) to changing climatic conditions (Dawson et al.,

2011; Pearson et al., 2014). For example, SDMs are increas-

ingly complemented with trait-based vulnerability assess-

ments (Heikkinen et al., 2009; Garcia et al., 2014a), while

the IUCN is using physiological, ecological, genetic and func-

tional characteristics, combined with simple measures of

exposure to climate change (e.g. Garcia et al., 2014b), to

ascribe new measures of vulnerability to thousands of species

for eventual inclusion in the Red List Categories and Criteria

(IUCN, 2011; Foden et al., 2013).

Through methods such as these, conservation planners may

embrace valuable insights provided by popular SDM method-

ologies, while addressing their inherent limitations with

respect to narrow-ranging species that have traditionally un-

derpinned global allocation of conservation funds. Effective

biodiversity conservation relies on our capacity to predict cli-

mate change impacts on all species, and not just those that

fulfil biologically arbitrary occurrence thresholds required by

one tool in the box. Our results suggest, however, that while

current priorities can be highly sensitive to the exclusion of

narrow-ranging taxa from SDM, likely strongholds for both

narrow- and wide-ranging species under projected climate

change are consistent with existing priority schemes. We con-

clude that while methods to tackle the rare species conun-

drum continue to advance, the focus on conserving habitat

connectivity and ecological functioning at these sites ought to

be maintained (Ricketts et al., 2005; Hodgson et al., 2009;

Iwamura et al., 2013; Pearson et al., 2014).

ACKNOWLEDGEMENTS

We thank many experts who contributed to the African

amphibian database, in particular through the IUCN Species

Survival Commission Specialist Group on African Herpetol-

ogy, and contributors to the Global Amphibian Assessment.

Additional distribution maps were provided by A. Schiøtz

(tree frogs), J. Poynton (bufonids), A. Channing (southern

African), S. Loader (Caecilians), M. R€odel (West Africa), M.

Menegon, K.M. Howell, D. Broadley and N. Doggart (East-

ern Arc and Coastal Forests). For access to the database,

please contact L.H. or C.R. at the University of Copenhagen.

P.J.P. was funded by the Ministry for Foreign Affairs of Fin-

land and by the Leverhulme Trust, R.A.G. by the Portuguese

Foundation for Science and Technology and by the European

Social Fund, W.F. by the IUCN Global Species Programme,

1316 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.

C.H. by LOEWE and N.D.B. by Copenhagen University,

WWF-US Conservation Science Program and UNEP-

WCMC. N.D.B., R.A.G., C.H. and C.R. acknowledge the sup-

port of the Danish National Research Foundation. Thank

you to Richard Pearson, David Nogu�es-Bravo, Jane Hill,

Anni Arponen, Jake Overton and an anonymous reviewer for

insightful feedback on earlier versions of the manuscript.

REFERENCES

Ahrends, A., Rahbek, C., Bulling, M.T., Burgess, N.D., Platts,

P.J., Lovett, J.C., Kindemba, V.W., Owen, N., Sallu, A.N.,

Marshall, A.R., Mhoro, B.E., Fanning, E. & Marchant, R.

(2011) Conservation and the botanist effect. Biological Con-

servation, 144, 131–140.

Ara�ujo, M.B. & Peterson, A.T. (2012) Uses and misuses of

bioclimatic envelope modeling. Ecological Society of Amer-

ica, 93, 1527–1539.

Balmford, A. & Gaston, K.J. (1999) Why biodiversity surveys

are good value. Nature, 398, 204–205.

Bateman, B.L., Murphy, H.T., Reside, A.E., Mokany, K. &

VanDerWal, J. (2013) Appropriateness of full-, partial- and

no-dispersal scenarios in climate change impact modelling.

Diversity and Distributions, 19, 1224–1234.

Bennie, J., Hodgson, J.A., Lawson, C.R., Holloway, C.T.R.,

Roy, D.B., Brereton, T., Thomas, C.D. & Wilson, R.J.

(2013) Range expansion through fragmented landscapes

under a variable climate. Ecology Letters, 16, 921–929.

Burgess, N., Butynski, T., Cordeiro, N., Doggart, N., Fjeldsa,

J., Howell, K., Kilahama, F., Loader, S., Lovett, J. & Mbil-

inyi, B. (2007) The biological importance of the Eastern

Arc Mountains of Tanzania and Kenya. Biological Conser-

vation, 134, 209–231.

Busby, J.R. (1991) BIOCLIM – a bioclimatic analysis and

prediction system. Nature conservation: cost effective biologi-

cal surveys and data analysis (ed. by C.R. Margules and

M.P. Austin), pp. 64–68. CSIRO, East Melbourne, Vic.,

Australia.

Butchart, S.H.M., Walpole, M., Collen, B. et al. (2010) Glo-

bal biodiversity: indicators of recent declines. Science, 328,

1164–1168.

Chessel, D., Dufour, A.B. & Thioulouse, J. (2004) The ade4

package-I-One-table methods. R News, 4, 5–10.

Da Fonseca, G.A.B., Balmford, A., Bibby, C., Boitani, L.,

Corsi, F., Brooks, T., Gascon, C., Olivieri, S., Mittermeier,

R.A., Burgess, N., Dinerstein, E., Olson, D., Hannah, L.,

Lovett, J., Moyer, D., Rahbek, C., Stuart, S. & Williams, P.

(2000) . . . following Africa’s lead in setting priorities. Nat-

ure, 405, 393–394.

Dawson, T.P., Jackson, S.T., House, J.I., Prentice, I.C. &

Mace, G.M. (2011) Beyond predictions: biodiversity con-

servation in a changing climate. Science, 332, 53–58.

De Klerk, H.M., Crowe, T.M., Fjelds�a, J. & Burgess, N.D.

(2002) Patterns of species richness and narrow endemism

of terrestrial bird species in the Afrotropical region. Journal

of Zoology, 256, 327–342.

Dol�edec, S., Chesse, D. & Gimaret-Carpentier, C. (2000)

Niche separation in community analysis: a new method.

Ecology, 81, 2914–2927.

Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G.,

Carr�e, G., Marqu�ez, J.R.G., Gruber, B., Lafourcade, B.,

Leit~ao, P.J., M€unkem€uller, T., McClean, C., Osborne, P.E.,

Reineking, B., Schr€oder, B., Skidmore, A.K., Zurell, D. &

Lautenbach, S. (2013) Collinearity: a review of methods to

deal with it and a simulation study evaluating their perfor-

mance. Ecography, 36, 27–46.

Early, R. & Sax, D.F. (2011) Analysis of climate paths reveals

potential limitations on species range shifts. Ecology Letters,

14, 1125–1133.

Elith, J., Graham, C.H., Anderson, R.P. et al. (2006) Novel

methods improve prediction of species’ distributions from

occurrence data. Ecography, 29, 129–151.

Feeley, K.J. & Silman, M.R. (2011) The data void in model-

ing current and future distributions of tropical species.

Global Change Biology, 17, 626–630.

Ferrier, S. & Guisan, A. (2006) Spatial modelling of biodiver-

sity at the community level. Journal of Applied Ecology, 43,

393–404.

Fjelds�a, J., Ehrlich, D., Lambin, E. & Prins, E. (1997) Are

biodiversity “hotspots” correlated with current ecoclimatic

stability? A pilot study using the NOAA-AVHRR remote

sensing data. Biodiversity and Conservation, 6, 401–422.

Foden, W.B., Butchart, S.H.M., Stuart, S.N., Vi�e, J.-C.,

Akc�akaya, H.R., Angulo, A., DeVantier, L.M., Gutsche, A.,

Turak, E., Cao, L., Donner, S.D., Katariya, V., Bernard, R.,

Holland, R.A., Hughes, A.F., O’Hanlon, S.E., Garnett, S.T.,

S�ekercio�glu, C� .H. & Mace, G.M. (2013) Identifying the

world’s most climate change vulnerable species: a system-

atic trait-based assessment of all birds, amphibians and

corals. PLoS ONE, 8, e65427.

Franklin, J. (2010) Mapping species distributions. Cambridge

University Press, Cambridge, UK.

Franklin, J. (2013) Species distribution models in conserva-

tion biogeography: developments and challenges. Diversity

and Distributions, 19, 1217–1223.

Franklin, J., Davis, F.W., Ikegami, M., Syphard, A.D., Flint,

L.E., Flint, A.L. & Hannah, L. (2013) Modeling plant spe-

cies distributions under future climates: how fine scale do

climate projections need to be? Global Change Biology, 19,

473–483.

Friedman, J.H. (1991) Multivariate adaptive regression

splines. Annals of Statistics, 19, 1–67.

Friedman, J.H. (2001) Greedy function approximation: a gra-

dient boosting machine. Annals of Statistics, 29, 1189–1232.

Garcia, R.A., Burgess, N.D., Cabeza, M., Rahbek, C. & Ara-

�ujo, M.B. (2012) Exploring consensus in 21st century pro-

jections of climatically suitable areas for African

vertebrates. Global Change Biology, 18, 1253–1269.

Garcia, R.A., Ara�ujo, M.B., Burgess, N.D., Foden, W.B., Gut-

sche, A., Rahbek, C. & Cabeza, M. (2014a) Matching spe-

cies traits to projected threats and opportunities from

climate change. Journal of Biogeography, 41, 724–735.

Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd 1317

Species distribution models and conservation priorities

Garcia, R.A., Cabeza, M., Rahbek, C. & Araujo, M.B. (2014b)

Multiple dimensions of climate change and their implica-

tions for biodiversity. Science, 344, 1247579.

Gaston, K.J. (1994) Rarity. Chapman and Hall, London, UK.

Gereau, R.E., Taylor, C.M., Bodine, S. & Kindeketa, W.J.

(2010) Plant conservation assessment in the Eastern Arc

Mountains and coastal forests of Tanzania and Kenya. Avail-

able at: http://www.mobot.org/MOBOT/Research/tanzania/

cepf.shtml/ (accessed 1 April 2014).

Golicher, D.J., Cayuela, L., Alkemade, J.R.M., Gonz�alez-

Espinosa, M. & Ram�ırez-Marcial, N. (2008) Applying

climatically associated species pools to the modelling of

compositional change in tropical montane forests. Global

Ecology and Biogeography, 17, 262–273.

Graham, C.H., Ferrier, S., Huettman, F., Moritz, C. & Peter-

son, A.T. (2004) New developments in museum-based

informatics and applications in biodiversity analysis. Trends

in Ecology & Evolution, 19, 497–503.

Hannah, L., Midgley, G.F., Lovejoy, T., Bond, W.J., Bush,

M., Lovett, J.C., Scott, D. & Woodward, F.I. (2002) Con-

servation of biodiversity in a changing climate. Conserva-

tion Biology, 16, 264–268.

Harrell, F.E., Lee, K.L., Califf, R.M., Pryor, D.B. & Rosati,

R.A. (1984) Regression modelling strategies for improved

prognostic prediction. Statistics in Medicine, 3, 143–152.

Heikkinen, R.K., Luoto, M., Leikola, N., P€oyry, J., Settele, J.,

Kudrna, O., Marmion, M., Fronzek, S. & Thuiller, W.

(2009) Assessing the vulnerability of European butterflies

to climate change using multiple criteria. Biodiversity and

Conservation, 19, 695–723.

Hernandez, P.A., Graham, C.H., Master, L.L. & Albert, D.L.

(2006) The effect of sample size and species characteristics

on performance of different species distribution modeling

methods. Ecography, 29, 773–785.

Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G. & Jarvis,

A. (2005) Very high resolution interpolated climate sur-

faces for global land areas. International Journal of Clima-

tology, 25, 1965–1978.

Hodgson, J.A., Thomas, C.D., Wintle, B.A. & Moilanen, A.

(2009) Climate change, connectivity and conservation deci-

sion making: back to basics. Journal of Applied Ecology, 46,

964–969.

Hof, C., Rahbek, C. & Ara�ujo, M.B. (2010) Phylogenetic sig-

nals in the climatic niches of the world’s amphibians. Ecog-

raphy, 33, 242–250.

Hof, C., Ara�ujo, M.B., Jetz, W. & Rahbek, C. (2011a)

Additive threats from pathogens, climate and land-use

change for global amphibian diversity. Nature, 480, 516–

519.

Hof, C., Levinsky, I., Ara�ujo, M.B. & Rahbek, C. (2011b)

Rethinking species’ ability to cope with rapid climate

change. Global Change Biology, 17, 2987–2990.

Hole, D.G., Willis, S.G., Pain, D.J., Fishpool, L.D., Butchart,

S.H.M., Collingham, Y.C., Rahbek, C. & Huntley, B. (2009)

Projected impacts of climate change on a continent-wide

protected area network. Ecology Letters, 12, 420–431.

Huntley, B., Collingham, Y.C., Green, R.E., Hilton, G.M.,

Rahbek, C. & Willis, S.G. (2006) Potential impacts of cli-

matic change upon geographical distributions of birds. Ibis,

148, 8–28.

IUCN (2011) Guidelines for Using The IUCN Red List Cate-

gories and Criteria. Version 9.0. Prepared by the IUCN

Standards and Petitions Subcommittee.

Iwamura, T., Guisan, A., Wilson, K.A. & Possingham, H.P.

(2013) How robust are global conservation priorities to cli-

mate change? Global Environmental Change, 23, 1277–1284.

Jetz, W. & Rahbek, C. (2002) Geographic range size and

determinants of avian species richness. Science, 297, 1548–

1551.

Kapos, V., Rhind, J., Edwards, M., Price, M.F. & Ravilious,

C. (2000) Developing a map of the world’s mountain for-

ests. Forests in sustainable mountain development: a state of

knowledge report for 2000 (ed. by M.F. Price and N. Butt),

pp. 4–9. CAB International, Wallingford.

K€uper, W., Sommer, J.H., Lovett, J.C. & Barthlott, W.

(2006) Deficiency in African plant distribution data – miss-

ing pieces of the puzzle. Botanical Journal of the Linnean

Society, 150, 355–368.

Liu, C., White, M. & Newell, G. (2013) Selecting thresholds

for the prediction of species occurrence with presence-only

data. Journal of Biogeography, 40, 778–789.

Loarie, S.R., Carter, B.E., Hayhoe, K., McMahon, S., Moe,

R., Knight, C.A. & Ackerly, D.D. (2008) Climate change

and the future of California’s endemic flora. PLoS ONE, 3,

e2502.

Loarie, S.R., Duffy, P.B., Hamilton, H., Asner, G.P., Field,

C.B. & Ackerly, D.D. (2009) The velocity of climate

change. Nature, 462, 1052–1055.

Lomba, A., Pellissier, L., Randin, C., Vicente, J., Moreira, F.,

Honrado, J. & Guisan, A. (2010) Overcoming the rare spe-

cies modelling paradox: a novel hierarchical framework

applied to an Iberian endemic plant. Biological Conserva-

tion, 143, 2647–2657.

Margules, C.R. & Pressey, R.L. (2000) Systematic conserva-

tion planning. Nature, 405, 243–253.

McClean, C.J., Lovett, J.C., K€uper, W., Hannah, L., Sommer,

J.H., Barthlott, W., Termansen, M., Smith, G.F., Tokumine,

S. & Taplin, J.R.D. (2005) African plant diversity and cli-

mate change. Annuals of the Missouri Botanical Garden, 92,

139–152.

Mittermeier, R.A., Gil, P.R., Hoffmann, M., Pilgrim, J., Brooks,

T., Mittermeier, C.G., Lamoreux, J. & da Fonseca, G.A.B.

(2004) Hotspots revisited: Earth’s biologically richest and most

endangered ecoregions. CEMEX, Mexico City, Mexico.

Moilanen, A., Wilson, K.A. & Possingham, H.P. (2009) Spa-

tial conservation prioritization. Oxford University Press,

Oxford.

Myers, N., Mittermeier, R.A., Mittermeier, C.G., Da Fonseca,

G.A. & Kent, J. (2000) Biodiversity hotspots for conserva-

tion priorities. Nature, 403, 853–858.

Ohlem€uller, R., Anderson, B.J., Ara�ujo, M.B., Butchart,

S.H.M., Kudrna, O., Ridgely, R.S. & Thomas, C.D. (2008)

1318 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.

The coincidence of climatic and species rarity: high risk to

small-range species from climate change. Biology Letters, 4,

568–572.

Olson, D.M. & Dinerstein, E. (1998) The Global 200: a rep-

resentation approach to conserving the Earth’ s most

biologically valuable ecoregions. Conservation Biology, 12,

502–515.

Ovaskainen, O. & Soininen, J. (2011) Making more out of

sparse data: hierarchical modeling of species communities.

Ecology, 92, 289–295.

Pearson, R.G. & Dawson, T.P. (2003) Predicting the impacts

of climate change on the distribution of species: are biocli-

mate envelope models useful? Global Ecology and Biogeog-

raphy, 12, 361–371.

Pearson, R.G., Stanton, J.C., Shoemaker, K.T., Aiello-lam-

mens, M.E., Ersts, P.J., Horning, N., Fordham, D.A., Rax-

worthy, C.J., Ryu, H.Y., Mcnees, J. & Akc�akaya, H.R.

(2014) Life history and spatial traits predict extinction

risk due to climate change. Nature Climate Change, 4,

217–221.

Phillips, S.J. & Dudik, M. (2008) Modeling of species distri-

butions with Maxent: new extensions and a comprehensive

evaluation. Ecography, 31, 161–175.

Pimm, S.L., Ayres, M., Balmford, A. et al. (2001) Can we

defy nature’s end? Science, 293, 2207–2208.

Platts, P.J., Ahrends, A., Gereau, R.E., McClean, C.J., Lov-

ett, J.C., Marshall, A.R., Pellikka, P.K.E., Mulligan, M.,

Fanning, E. & Marchant, R. (2010) Can distribution

models help refine inventory-based estimates of conserva-

tion priority? A case study in the Eastern Arc forests of

Tanzania and Kenya. Diversity and Distributions, 16, 628–

642.

Platts, P.J., Gereau, R.E., Burgess, N.D. & Marchant, R.

(2013) Spatial heterogeneity of climate change in an Afro-

montane centre of endemism. Ecography, 36, 518–530.

Platts, P.J., Omeny, P.A. & Marchant, R. (2014) AFRICLIM:

high-resolution climate projections for ecological applica-

tions in Africa. African Journal of Ecology. doi:10.1111/aje.

12180

Pressey, R.L., Hager, T.C., Ryan, K.M., Schwarz, J., Wall, S.,

Ferrier, S. & Creaser, P.M. (2000) Using abiotic data for

conservation assessments over extensive regions: quantita-

tive methods applied across New South Wales, Australia.

Biological Conservation, 96, 55–82.

Pyron, R.A. & Wiens, J.J. (2013) Large-scale phylogenetic

analyses reveal the causes of high tropical amphibian diver-

sity. Proceedings of the Royal Society B, 280, 1471–2954.

Randin, C.F., Engler, R., Normand, S., Zappa, M., Zimmer-

mann, N.E., Pearman, P.B., Vittoz, P., Thuiller, W. & Gui-

san, A. (2009) Climate change and plant distribution: local

models predict high-elevation persistence. Global Change

Biology, 15, 1557–1569.

R-Core-Team (2012) R: a language and environment for sta-

tistical computing. R Foundation for Statistical Computing,

Vienna. Available at: http://www.r-project.org/ (version

3.0.2).

Ricketts, T.H., Dinerstein, E., Boucher, T. et al. (2005) Pin-

pointing and preventing imminent extinctions. Proceedings

of the National Academy of Sciences USA, 102, 18497–18501.

Sandel, B., Arge, L., Dalsgaard, B., Davies, R.G., Gaston, K.J.,

Sutherland, W.J. & Svenning, J.-C. (2011) The influence of

late quaternary climate-change velocity on species ende-

mism. Science, 334, 660–664.

Shoo, L.P., Storlie, C., Vanderwal, J., Little, J. & Williams,

S.E. (2011) Targeted protection and restoration to conserve

tropical biodiversity in a warming world. Global Change

Biology, 17, 186–193.

Sodhi, N.S., Bickford, D., Diesmos, A.C., Lee, T.M., Koh,

L.P., Brook, B.W., Sekercioglu, C.H. & Bradshaw, C.J.A.

(2008) Measuring the meltdown: drivers of global amphib-

ian extinction and decline. PLoS ONE, 3, e1636.

Stattersfield, A., Crosby, M.J., Long, A.J. & Wege, D.C.

(1998) Endemic Bird Areas of the world: priorities for biodi-

versity conservation. BirdLife International, Cambridge, UK.

Stockwell, D. & Peters, D. (1999) The GARP modelling sys-

tem: problems and solutions to automated spatial predic-

tion. International Journal of Geographical Information

Science, 13, 143–158.

Stockwell, D.R. & Peterson, A.T. (2002) Effects of sample size

on accuracy of species distribution models. Ecological Mod-

elling, 148, 1–13.

Trivedi, M.R., Berry, P.M., Morecroft, M.D. & Dawson, T.P.

(2008) Spatial scale affects bioclimate model projections of

climate change impacts on mountain plants. Global Change

Biology, 14, 1089–1103.

Walker, P.A. & Cocks, K.D. (1991) HABITAT – a procedure

for modeling a disjoint environmental envelope for a plant

or animal species. Global Ecology and Biogeography Letters,

1, 108–118.

Walsh, R.P.D. & Lawler, D.M. (1981) Rainfall seasonality:

description, spatial patterns and change through time.

Weather, 36, 201–208.

Wiens, J.A. & Bachelet, D. (2010) Matching the multiple

scales of conservation with the multiple scales of climate

change. Conservation Biology, 24, 51–62.

Wiens, J.J., Graham, C.H., Moen, D.S., Smith, S.A. & Reeder,

T.W. (2006) Evolutionary and ecological causes of the lati-

tudinal diversity gradient in hylid frogs: treefrog trees

unearth the roots of high tropical diversity. The American

Naturalist, 168, 579–596.

Wiens, J.A., Stralberg, D., Jongsomjit, D., Howell, C.A. &

Snyder, M.A. (2009) Niches, models, and climate change:

assessing the assumptions and uncertainties. Proceedings of

the National Academy of Sciences USA, 106(Suppl), 19729–

19736.

Williams, P., Hannah, L., Andelman, S., Midgley, G., Ara�ujo,

M., Hughes, G., Manne, L., Martinez-Meyer, E. & Pearson,

R. (2005) Planning for climate change: identifying mini-

mum-dispersal corridors for the Cape Proteaceae. Conser-

vation Biology, 19, 1063–1074.

Williams, J.W., Jackson, S.T. & Kutzbach, J.E. (2007) Pro-

jected distributions of novel and disappearing climates by

Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd 1319

Species distribution models and conservation priorities

2100 AD. Proceedings of the National Academy of Sciences

USA, 104, 5738–5742.

Williams, J.N., Seo, C., Thorne, J., Nelson, J.K., Erwin, S.,

O’Brien, J.M. & Schwartz, M.W. (2009) Using species dis-

tribution models to predict new occurrences for rare

plants. Diversity and Distributions, 15, 565–576.

Wilson, K.A., McBride, M.F., Bode, M. & Possingham, H.P.

(2006) Prioritizing global conservation efforts. Nature, 440,

337–340.

Wisz, M.S., Hijmans, R.J., Li, J., Peterson, A.T., Graham,

C.H. & Guisan, A. (2008) Effects of sample size on the per-

formance of species distribution models. Diversity and Dis-

tributions, 14, 763–773.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the

online version of this article:

Figure S1 Comparison of Outlying Mean Index results for

eligible versus omitted species.

Figure S2 Future climate anomalies and novel climates.

Figure S3 Variable contributions in MAXENT models.

Figure S4 Estimates of species richness and range-size rarity

based on the eligible species set, compared for MAXENT pre-

dictions versus multidimensional niche envelopes.

Table S1 African amphibians included in the study, includ-

ing range size and Red List status.

Table S2 Summary of IUCN Red List status for all amphib-

ian species included in the study, detailed separately for eligi-

ble and omitted species.

Table S3 Environmental variable loadings and variance

explained by the two-first axes of the Outlying Mean Index

Analysis, on present-day climate and on future climate

anomalies.

BIOSKETCH

Philip J. Platts is a Research Fellow at the University of

York, UK. His recent work has focused on mountain forests

in eastern Africa, their biological importance and the ecosys-

tem services they provide.

Author contributions: N.D.B. and C.H. conceived the idea;

P.J.P. and R.A.G. designed and implemented the analyses;

L.H. and W.F. worked on the amphibian database; P.J.P.,

R.A.G. and N.D.B. led the writing; and all authors inter-

preted the results and commented on the text.

Editor: Matt Fitzpatrick

1320 Diversity and Distributions, 20, 1307–1320, ª 2014 John Wiley & Sons Ltd

P. J. Platts et al.